Have just posted to WUWT the following on global temperature anomalies:-

Thanks Luboš for a well-thought out article, and nicely summarised by

“The “error of the measurement” of the warming trend is 3 times larger than the result!”

One of the implications of this wide variability, and the concentration of temperature measurements in a small proportion of the land mass (with very little from the oceans covering 70% of the globe) is that one must be very careful in the interpretation of the data. Even if the surface stations were totally representative and uniformly accurate (no UHI) and the raw data properly adjusted (Remember Darwin, Australia on this blog?), there are still normative judgements to be made to achieve a figure.

I have done some (much cruder) analysis comparing HADCRUT3 to GISSTEMP for the period 1880 to 2010, which helps illustrate these judgemental decisions.

1. The temperature series agree on the large fluctuations, with the exception of the post 1945 cooling – it happens 2 or 3 years later and more slowly in GISSTEMP.

2. One would expect greater agreement with recent data in more recent years. But since 1997 the difference in temperature anomalies has widened by nearly 0.3 celsius – GISSTEMP showing rapid warming and HADCRUT showing none.

3. If you take the absolute change in anomaly from month to month and average from 1880 to 2010, GISSTEMP is nearly double that of HADCRUT3 – 0.15 degrees v 0.08. The divergence in volatility reduced from 1880 to the middle of last century, when GISSTEMP was around 40% more volatile than HADCRUT3. But since then the relative volatility has increased. The figures for the last five years are respectively about 0.12 and 0.05 degrees. That is GISSTEMP is around 120% more volatile that HADCRUT3.

This all indicates that there must be greater clarity in the figures. We need the temperature indices to be compiled by qualified independent statisticians, not by those who major in another subject. This is particularly true of the major measure of global warming, where there is more than a modicum of partisan elements.

These graphs help illustrate the points made. Please note that I use overlapping moving averages, so it is for illustrative purposes only.

NB. Luboš Motl’s article was cross-posted from his blog here